We will be undergoing planned maintenance on Oct 7th 6:00AM UTC / Oct 7th 2:00AM ET

DebugDojo

DebugDogo is a teacher debugger that tell students what is wrong with their code and help them like a teacher

Comment

Inspiration As beginners, we noticed most AI coding tools just hand you the fixed code, so you copy-paste and never actually learn to debug. We wanted an AI tutor that does the opposite: it asks questions, gives hints, and only shows the fix when you're truly stuck so you build real debugging skills, not copy-paste habits.

What it does DebugDogo lets you paste or write buggy code in 6 languages (C++, C, Java, Python, JavaScript, TypeScript). The AI tutor diagnoses the bug, asks a guided question instead of revealing the fix, and gives progressive hints if you're stuck. While you type, it even leaves friendly inline comments pointing at suspicious lines. If your code is already correct, it congratulates you and shows the output instead of forcing you through unnecessary steps. You can also Run your code to see simulated terminal output at any time.

How we built it Frontend: React + Vite + Tailwind CSS, with Monaco Editor for a real code-editing experience. Backend: Node.js + Express, deployed as a serverless function on Vercel. AI: Groq API running openai/gpt-oss-120b for both the tutoring logic and the live inline code review. Both frontend and backend are deployed separately on Vercel and connected via environment variables . Challenges we ran into Getting the frontend and backend talking correctly after deployment was the hardest part — CORS errors and mismatched API URLs between Vercel preview and production deployments took real debugging (ironic, for a debugging app!). We also had to tune the AI prompt carefully so it never leaked the solution early, and optimized the live inline-comment feature to respond fast without breaking the typing experience.

Accomplishments that we're proud of Building a full AI tutoring flow (diagnose → question → hint → fix) that actually teaches instead of just answering. We're also proud of the live AI code comments while typing, which felt like a genuinely original feature for a beginner tool.

What we learned We learned how to design AI prompts that hold back information on purpose, how to deploy a full-stack app split across two Vercel projects, and a lot about real-world CORS and environment variable debugging.

What's next for DebugDogo

Adding a real sandboxed code execution engine instead of AI-simulated output, a difficulty mode for intermediate learners, a bug library covering more concept categories and many more to come

Built With

Share this project:

Updates

Submission history